October 29th, 2026
Me on my Alaska trip in 2025
We’ve all been told that AI is the ultimate shortcut, a magic box where you type a casual question and get an expert answer. But anyone who has tried to use it for complex, real-world projects knows the truth: if you treat AI like a basic search engine, it will confidently feed you outdated software steps, miss critical local laws, and lead you straight into a technical dead end. Over the past three months, I’ve used Google AI as a daily partner to manage everything from international travel logistics to a complex website migration. In the process, I learned that getting elite results isn't about finding the perfect prompt—it’s about knowing how to actively manage your digital assistant, step back when you're trapped in the weeds, and avoid the devastating trap of "rat-holing."
Here are the 8 master rules I developed to turn AI from a basic search box into an elite assistant:
1. Bookmark Your Threads, but Don't Trust Their Memory
AI does a remarkable job of maintaining a long conversation, provided you bookmark the exact page of your thread and keep returning to it. However, do not assume it remembers everything flawlessly. It can suffer from "context drift" or offer completely contradictory advice on different days.
The Fix: Keep a separate running note (I use Evernote) of key decisions. If the AI seems confused, explicitly remind it: "We discussed this yesterday, and we decided to go with Option B. Let's restart from there." This instantly snaps its focus back into place.
2. The Travel Trap: Look for What’s Missing
When planning international travel, AI is invaluable for navigating flight rules, luggage weight limits, health vaccinations, visas, and customs. However, never assume a generated itinerary is entirely safe. AI frequently leaves out critical, hyper-local updates.
For example, it initially missed a strict, relatively new law in Tanzania: civilian camo gear is illegal and will be confiscated at customs. Always use AI to build the framework, but manually verify the final compliance details.
3. Stop Asking Questions; Start Specifying Variables
Treating AI like a basic Google search will yield frustrating, generic results. To get precise answers, you must treat your prompt like a set of constraints. You must explicitly specify:
The exact software version: If you don't tell it you are on the newest version of Squarespace, it will feed you step-by-step navigation instructions for a version from five years ago.
The time of year: Essential for accurate weather, packing lists, and wildlife behavior forecasts.
Live data overrides: AI relies heavily on cached web information. For product research like SSD drives, cached data means quoted prices are almost always wrong. You must explicitly command it: "Search the web right now for live, current pricing on this exact model."
Failing to do this can lead to massive financial surprises—such as an AI quoting a platform cost of $100, only for you to find out later that the features you requested actually push you into a $300 tier.
4. The Power of "Are You Sure?"
The single most disruptive trick I learned is adding the phrase "Are you sure?" to my queries. Because AI models operate on probability, they will occasionally present unverified information with complete confidence.
When I noticed an anomaly and asked "Are you sure?", the AI would trigger a self-correction loop, catch its own mistake, and provide a significantly more accurate, reality-tested alternative. Make this a standard step before acting on critical advice.
5. Don’t Rat-Hole: Think Bigger Than the Error
When trying to migrate my historical blog posts and images from Squarespace to Ghost, I spent an entire frustrating day trying to force an XML export to behave. The AI and I completely "rat-holed"—getting trapped in a loop of writing endless custom CSS and Javascript tweaks inside Ghost to patch up a broken import. It was a total failure.
The breakthrough came when I stepped away, thought about the larger data architecture overnight, and changed my strategy. Instead of forcing Ghost to accept a messy file, I asked the AI to write a custom Python script to clean, format, and massage the raw Squarespace data before uploading it.
The lesson? When you find yourself stuck in the weeds adjusting code patches with an AI, step back, look at the big picture, and ask the AI to build a tool to fix the data source instead.
6. Give the AI a Persona or "Role" to Shift Its Perspective
You aren't just asking a generic database for assistance; you get drastically better answers when you explicitly tell the AI what filter to use. When shifting from writing a raw field story to structuring a storefront entry, you should change the AI's role from a naturalist storyteller to a conversion-focused e-commerce specialist.
Don't just ask "how do I sell these prints?" Tell the AI: "Act as an expert art e-commerce consultant. How should I bundle these three images to maximize my total order value?" Swapping roles forces the AI to pull from different, more relevant pools of data.
7. Provide Real-Time Visual Anchors for Abstract Coding
Software platforms update their dashboards constantly, meaning textual navigation steps change on a whim. Trying to describe a screen layout in text can lead to loops of confusion. If a step-by-step technical instruction fails, don't keep guessing.
Pivot immediately and tell the AI exactly what your eyes see from left to right (e.g., "Save and Exit are on the left" or "I see all three images but they are cut off"). Describing your precise visual UI constraints breaks the AI out of standard template responses and lets it give you tailored, platform-specific hotfixes.
8. Exploit the "Incremental Building" Strategy
When tackling massive tasks, humans tend to want the final product delivered all at once. But with an AI, trying to do everything in one massive prompt usually results in an unstable, muddy output. Use a modular approach instead.
Break your large projects down into tight, independent milestones. Let the AI finish the story text first. Once you approve that, move to the image captions. Once that is locked, move to the shop pricing. Building incrementally keeps the AI's working memory laser-focused and prevents data clutter.